An Associated-Memory-Based Stock Price Predictor

摘要

We propose a novel method to predict stock price based on the Neural Associative Memory with Self-Organizing and Incremental Neural Networks (SOINN-AM). Our method has two advantages: 1) the predictor can determine its inner state space by the input training patterns automatically, 2) the predictor can modify itself by online-learning. Consequently, the predictor is more flexible for real world data than previous prediciton approaches. We demonstrate effectiveness of our approach with experiment result on real stock price data from the US and Japan market in 2002 - 2004.